Dynamic mode decomposition analysis of brain dynamics in autism spectrum disorder patients

Autism spectrum disorder (ASD) has been associated with atypical large-scale brain organization, yet most functional magnetic resonance imaging (fMRI) studies rely on static connectivity measures that do not explicitly characterize temporal dynamics. Here, we applied dynamic mode decomposition (DMD), a data-driven method that captures recurrent spatiotemporal patterns in terms of temporal persistence and oscillatory timing, to resting-state fMRI data from the Autism Brain Imaging Data Exchange (ABIDE; N = 849). Using a group-level DMD framework with subject-level mode estimation, we identified dynamic modes whose temporal properties differed between individuals with ASD and typically developing (TD) controls. In particular, ASD showed altered oscillatory timing in a posterior visual-parietal-temporal mode, and age-related associations with DMD features differed between ASD and TD groups, suggesting atypical developmental trajectories of large-scale temporal organization. Within the ASD group, DMD features were additionally associated with individual differences in IQ, indicating that temporal brain dynamics partially reflect cognitive heterogeneity in ASD. Spatiotemporal reconstruction and Neurosynth-based spatial correspondence analyses provided descriptive functional context for the extracted modes. Together, these findings suggest that DMD offers a compact framework for characterizing temporal organization of intrinsic brain activity and may capture dimensions of ASD-related neurocognitive variability beyond static connectivity alone. Oscillatory timing here refers to recurrence properties of low-frequency BOLD-derived dynamic modes rather than electrophysiological oscillations measured directly from neural signals.